23 Jul
2009
23 Jul
'09
11:45 p.m.
Hi,
I am completely in favor of dealing with the packages. I just disagree
a bit with the strategy :).
I believe that a package that includes other packages does not really
scale with the current tool support. Until the tool support is there,
we are much better off with using just scripts that load all the
needed parts.
From another point of view, I know it is a bit difficult to always
load the complete Moose but splitting the image will just lead to
pains when it will come to reporting issues due to different
configurations and due to the community being not so large at the
moment.
As I announced before I started to work on a Moose release. I am still
waiting for people to join if they want :). This can be a starting
point for people to build on. Ideally we could get one in about 2
weeks time. Not a perfect one, but a working one. And then we build on
that. This release will always be loadable on a Pharo-Core with
MooseLoader load.
Of course, if people have very specific needs, we can provide scripts
for that, but I would not want to have everyone use a different
configuration. This will also put pressure on the main release to be
small and not have redundancies. This process worked quite well in the
VW version.
Cheers,
Doru
On 23 Jul 2009, at 17:42, Simon Denier wrote:
I wonder if we could not refactor the dependencies
between the
required packages in Moose-All. Because it starts to look like a big
ball of mud. Granted we will soon have the tools (with Jannik) to
get a better comprehension of that but here it is:
Moose-all required packages:
Moose-Core-jannik_laval.128,
Moose-GenericImporter-simon_denier.laval.26,
Moose-LAN-jannik_laval.8, <========= test and resources
Moose-SmalltalkImporter-jannik_laval.ducasse.58,
Moose-ModelTest-simon_denier.26, <========= test and resources
Moose-ConformityStrategies-simon_denier.9,
Famix-Core-Alexandre_Bergel.84,
Famix-Implementation-simon_denier.37,
Famix-Smalltalk-simon_denier.23,
Famix-File-Alexandre_Bergel.15,
EMOF-stephane.ducasse.1, <======= this one should be removed
Moose-Test-PackageTwo-simon_denier.ducasse.3, <========= test and
resources
Moose-ReferenceModel-stephane_ducasse.ducasse.15, <========= test
and resources
Moose-Test-PackageOne-stephane.ducasse.2, <========= test and
resources
Famix-SmalltalkImporter-simon_denier.81,
Famix-LANTests-simon_denier.ducasse.22, <========= test and resources
CollectionExtensions-simon_denier.14,
Famix-Extensions-simon_denier.42,
Moose-File-Alexandre_Bergel.9, <========= test and resources
Moose-CookFamix3-simon_denier.18,
Moose-OBBrowser-simon_denier.50,
Moose-MondrianScripts-simon_denier.32, <==== require
MondrianExtensions and Mondrian
Moose-Hismo-simon_denier.12,
Famix-Test-simon_denier.4, <========= test and resources
Famix-SourceAnchor-tg.5,
Moose-MonticelloImporter-Alexandre_Bergel.7,
Moose-Finder-tg.28, <========= needs Glamour project
Moose also needs (but does not 'require') Fame project, Nile, Smacc,
RB, Omnibrowser, RIO for some extensions, and perhaps other things
we are not really aware of...
This is a lot...
In line with the current effort to better the MooseLoader, here is
my proposal for refactoring moose dependencies.
I distinguish between:
- external dependencies to external projects (Fame, Mondrian,
Glamour...) - managed by MooseLoader load scripts
- internal dependencies aka required packages for a single project -
managed by Monticello internal mechanisms
- hidden dependencies, which should be removed as soon as
identified :)
*Moose-Basic*
The goal is to have a minimal footprint moose image, for performing
batch operations, importing big projects, exporting big MSE, etc,
without relying on a browser.
It can also serve as a basis for people who want to customize their
moose image (like Jannik or I do, loading DSM after loading moose...)
external: Nile, Fame, Smacc (to be removed?), RBSmallDictionary
internal:
Moose-Core-jannik_laval.128,
Moose-GenericImporter-simon_denier.laval.26,
Moose-SmalltalkImporter-jannik_laval.ducasse.58,
Moose-ConformityStrategies-simon_denier.9,
Famix-Core-Alexandre_Bergel.84,
Famix-Implementation-simon_denier.37,
Famix-Smalltalk-simon_denier.23,
Famix-SmalltalkImporter-simon_denier.81,
CollectionExtensions-simon_denier.14,
Famix-Extensions-simon_denier.42,
Famix-SourceAnchor-tg.5,
borderline:
Famix-File-Alexandre_Bergel.15,
Moose-CookFamix3-simon_denier.18,
*Moose-Basic-Tests* (actually tests and resources package require a
refactoring of their own, as some are tangled and others are rather
empty)
internal:
Moose-Basic
Moose-LAN-jannik_laval.8, <========= test and resources
Moose-ModelTest-simon_denier.26, <========= test and resources
Moose-Test-PackageTwo-simon_denier.ducasse.3, <========= test and
resources
Moose-ReferenceModel-stephane_ducasse.ducasse.15, <========= test
and resources
Moose-Test-PackageOne-stephane.ducasse.2, <========= test and
resources
Famix-LANTests-simon_denier.ducasse.22, <========= test and resources
Moose-File-Alexandre_Bergel.9, <========= test and resources
Famix-Test-simon_denier.4, <========= test and resources
*Moose-All*
A complete distribution of Moose-related stable things for most usage.
external: Mondrian, Glamour, Omnibrowser (at least for know), RIO...
internal:
Moose-Basic
(Moose-Basic-Tests)
Moose-OBBrowser-simon_denier.50,
Moose-MondrianScripts-simon_denier.32, <==== require
MondrianExtensions and Mondrian
Moose-Hismo-simon_denier.12,
Moose-MonticelloImporter-Alexandre_Bergel.7,
Moose-Finder-tg.28, <========= needs Glamour project
Anyone has something to say?
--
Simon
_______________________________________________
Moose-dev mailing list
Moose-dev(a)iam.unibe.ch
https://www.iam.unibe.ch/mailman/listinfo/moose-dev
--
www.tudorgirba.com
"Some battles are better lost than fought."